Search failures: LLM-curated reading lists bridging libraries and discovery

Executive Summary

Search failures are a persistent problem for the roughly 30 million knowledge workers, graduate students, and researchers who together represent a $24.0B annual market; they spend significant time on fragmented discovery workflows that miss relevant papers, replicate searches, and produce overwhelming low-signal results. These users confront brittle search interfaces, paywalled silos, and the cognitive load of converting vague research questions into prioritized, learning-oriented reading plans. A practical product would use modern LLM comprehension to translate short prompts into multi-stage, personalized reading paths that aggregate library catalogs, preprints, citation graphs, and publisher metadata, with explicit provenance and exportable bibliographies. Core features should include grounding via open metadata/APIs, institutional authentication for access, human-in-the-loop curation for quality guarantees, and analytics that demonstrate time-to-insight and learning progress. The timing is attractive because three converging trends—LLMs that infer intent from sparse prompts, improving open metadata/APIs, and rising demand for personalized learning journeys—lower technical and adoption barriers; in our scoring the market is 92/100 with revenue potential at 88/100. With an estimated $800 average annual spend per user across 30M potential buyers, a focused B2B2C approach beginning with institutional pilots could scale to meaningful ARR if integration and retention are managed. To stand out you must build rigorous grounding and attribution, prioritize deep library and discovery integrations plus human oversight to reduce hallucination and licensing risk, and accept that competition is medium and the main challenges are earning institutional trust, handling API complexity, and proving measurable impact on time-to-insight before scaling.